Digital Signal Processing Reference
In-Depth Information
0
−20
−40
−60
Ω
Ω
c
=
π
/ 2
π
2
π
0
1
Ω
k
0
1 2 3 4 5 6 6 7 8 9 10
16
24
32
0
Ω
s
/2 =
π
Ω
s
= 2
π
0
−20
−40
Ω
2
π
−60
Ω
c
π
0
1
Ω
k
0
1 2 3 4 5 6 6 7 8 9 10
16
24
32
0
Ω
s
/2 =
π
Ω
s
= 2
π
Fig. 2.25 FIR filter design using frequency-sampling method. Above without a transition sam-
ple. Below with two transition samples (dotted curve for the first case without transition samples)
This command yields the impulse response of a length-N (i.e., order N - 1)
linear phase FIR filter with desired magnitude response, given by the vectors
Fd
and
Ad
.
Fd
is a vector of normalized frequency points, ranging from 0 to 1
(corresponding to the range 0 \ f \ f
s
/2 Hz), and
Ad
contains the desired mag-
nitudes for points in
Fd
. Example: To design an optimal FIR filter in MATLAB
with the specifications in the previous example, one can use:
h
¼
remez
ð
N
1
;½
00
:
50
:
61
;½
1100
Þ
Figure (
2.26
) shows the actual frequency response as compared to the ad-hoc
frequency-sampling approach described in
Sect. 2.6.5.2-A
.
Search WWH ::
Custom Search